For years the arrangement between search engines and websites was simple: the engine showed a link, a person clicked it, and the site got a visit. Clicks became the proof that search was working.
Zero-click search is what happens when that last step never takes place. AI assistants have made it far more common, and the uncomfortable consequence is that a report built on clicks can no longer tell you whether you are winning.
What is zero-click search?
Zero-click search is a search or AI answer that resolves the question without a click to any website. The person gets what they needed on the page or in the answer itself, and leaves.
The term first described traditional results pages. A question about the weather, a unit conversion or a business’s opening hours could be answered directly by a snippet or information panel, so no site received a visit.
From the searcher’s side, this is simply a faster answer. The effect lands on the sites whose information made the answer possible, and on anyone who relied on clicks as their measure of success.
For a long time that effect was limited to quick factual lookups, where few brands expected much traffic anyway. Questions that involved a purchase, a comparison or a judgment still sent people to websites, because a snippet could not do the weighing for them. That boundary is the one AI answers have moved.
AI answers extend zero-click from facts to decisions
AI assistants push zero-click behavior beyond simple facts into complex, considered questions. They can answer “which tool should I use” as completely as a results page answers “what time is it in Tokyo.”
A single response from an answer engine can:
- Explain what a category of product does and who it suits.
- Name several options and describe how they differ.
- Summarize what reviewers and users tend to say about each.
- Recommend a shortlist based on the constraints the buyer gave.
- Handle follow-up questions about pricing, integrations or alternatives.
That is much of the research a buyer used to do by opening tabs. When the assistant cites sources, some people click through to check. Many do not, because the answer already gave them what they came for. The brands named in that answer have shaped the decision whether or not anyone visited their site.
Why do clicks undercount influence?
Clicks only record the moment someone follows a link. They miss every case where an answer shaped a decision that was acted on later, somewhere else.
The influence happens before the visit
A buyer asks an assistant for the best options in a category. Your brand is named with a short, favorable description. The buyer does not click, but you are now on the shortlist, and everything that follows starts from there.
The visit arrives under another label
Days later the same buyer types your brand name into a browser, searches for it by name, or mentions it to a colleague who books the demo. Your analytics record a direct visit or a branded search. The answer that put you in consideration does not appear anywhere.
Losses leave no trace at all
If a competitor was named instead of you, there is no lost click to see. You simply never entered the conversation, and no report built on traffic can show it.
Falling traffic can hide rising influence
The same mechanism can make a healthy position look like decline. If more buyers get their answer from an assistant that recommends you, fewer of them need to visit your comparison pages or blog posts. Traffic to those pages can soften while your place on shortlists improves, and a clicks-only report would read that as a problem.
This is why referral data tends to understate the role of AI answers, and why finance teams reasonably ask for a clearer case. How to build one is covered in how to report AEO to a CFO.
What should brands measure instead of clicks?
Measure the answers themselves: whether they name you, how they describe you, which sources they rely on, and the visits that do arrive from assistants. Clicks remain a signal, but not the only one.
Whether answers name you
Start with visibility: across a fixed set of prompts that real buyers ask, the share of answers that mention your brand at all. Track it per engine, because ChatGPT, Gemini, Perplexity, Claude and Grok often answer the same question with different brands.
How answers describe you
Being named is only the first test. Read the description for:
- Position. Whether you are the first recommendation or a name at the end of a list.
- Share of voice. Your share of all brand mentions in those answers, compared with competitors, explained in what AI share of voice is.
- Sentiment. Whether the description is positive, neutral or hedged with caveats.
- Accuracy. Whether pricing, features and positioning are stated correctly.
Which sources answers cite
When engines retrieve pages, the cited URLs show which sites are shaping answers in your category. Some will be yours; many will be review sites, comparison articles, documentation or community threads. Those sources tell you where to focus, and they are measurable even when nobody clicks them.
The visits that do arrive
Some buyers do click, and AI-referred visits are worth tracking as a directional signal. Server logs add a second view: which AI crawlers read your pages, and which pages are crawled but never cited. Noma’s Agent Analytics reads those logs, and pairs them with answer-level metrics across five engines; it does not track conversions or revenue, so keep your existing analytics for that part.
What to do first
In rough order of return on effort:
- Add answer-level metrics next to traffic in your reporting, so stakeholders can see influence that clicks cannot show.
- Build a stable set of category, comparison and problem prompts, and measure against it consistently on every engine.
- Make your key pages quotable, with direct answers, clear headings and current facts, so an engine retrieving them summarizes you correctly.
- Find the sources engines cite for your category, and invest in being accurately described there.
Zero-click search is not a traffic problem to be solved. It is a change in where opinions form, and the brands that measure those places will see what the others cannot.